Wasserstein Distributionally Robust Optimization with Wasserstein Barycenters

TL;DR

This paper introduces a novel approach for distributionally robust optimization using Wasserstein barycenters to aggregate data from multiple sources, providing a tractable formulation with strong theoretical guarantees and improved estimation performance.

Contribution

It proposes using Wasserstein barycenters to construct nominal distributions in distributionally robust optimization with a tractable convex reformulation and theoretical guarantees.

Findings

Outperforms existing estimators in Gaussian inverse covariance estimation

Provides finite-sample and asymptotic guarantees

Demonstrates effectiveness in low- and high-dimensional regimes

Abstract

In many applications in statistics and machine learning, the availability of data samples from multiple possibly heterogeneous sources has become increasingly prevalent. On the other hand, in distributionally robust optimization, we seek data-driven decisions which perform well under the most adverse distribution from a nominal distribution constructed from data samples within a certain discrepancy of probability distributions. However, it remains unclear how to achieve such distributional robustness in model learning and estimation when data samples from multiple sources are available. In this work, we propose constructing the nominal distribution in optimal transport-based distributionally robust optimization problems through the notion of Wasserstein barycenter as an aggregation of data samples from multiple sources. Under specific choices of the loss function, the proposed formulation admits a tractable reformulation as a finite convex program, with powerful finite-sample and asymptotic guarantees. As an illustrative example, we demonstrate with the problem of distributionally robust sparse inverse covariance matrix estimation for zero-mean Gaussian random vectors that our proposed scheme outperforms other widely used estimators in both the low- and high-dimensional regimes.